Freelance AI Engineer · Stuttgart · Remote across Germany
AI that actually runs in your company.
I’m Niclas. I build knowledge assistants, AI agents and automations that work with your data and systems – from the first process check to stable operation. For large enterprises and mid-sized companies alike.
Enterprise
Multi-agent system for a German automotive group
In production
RAG assistant in daily use
Leadership
Former CTO with equity
Business
Economist & ex-KPMG
01 — Starting point
Sound familiar?
- A
“We want a chatbot for our internal wiki.”
- B
“This process eats up hours every week. Surely AI can do that.”
- C
“We need AI to cut the workload in support.”
- D
“Our prototype looks good, but nobody dares to take it live.”
The idea is rarely the problem. The hard part comes after: Which data may the AI see? Who checks whether the answers are right? Who takes care of it when the model changes? And does it pay off at all?
That’s exactly where I come in.
02 — Services
What I build for you
Knowledge Assistants
Your knowledge from wikis, PDFs and manuals as a chat that answers with sources.
How I build knowledge assistantsAI Agents
Systems that combine several data sources and tools and handle complex questions on their own.
How I build AI agentsAI Automation
Recurring workflows where AI takes over exactly the step where fixed rules fail.
How I automate processesOperations & Continuous Improvement
Monitoring, quality measurement and updates, so your system is as good in a year as it is today.
How I look after systems03 — Projects
AI systems in use today
- 01
Multi-agent system for a German automotive group
Specialized agents answer analysis questions across several data sources, in a chat interface with live streaming.
Enterprise · Multi-agent · Databricks
- 02
AI assistant for the knowledge of a school
An internal wiki and a legal reference work of around 1,400 pages, searchable as a chat with sources. Built alone, from requirements to operation.
RAG · In production · Solo
- 03
Location analysis for a self-storage provider
Location research cut from several days to a few minutes, and data-tool subscriptions worth several thousand euros per year replaced.
Mid-sized company · AI analysis
04 — How I work
From process to operation
What you get at the end is not a slide deck, but a system your people use.
- 01
Understand
I get to know your process and check whether it needs AI at all.
- 02
Prove
A small pilot with real data shows what works.
- 03
Productionize
Integration, permissions, tests and monitoring.
- 04
Operate
I stay until it runs. And after that, too.
05 — Why me
Why companies work with me
- 01
Business first, then code.
Economist, formerly an auditor at KPMG, later CTO with equity. I ask about benefit and cost before I ask about the model.
- 02
The whole system.
Data, backend, interface, permissions, tests, deployment. No prompt prototype that someone else has to rescue later.
- 03
Enterprise and mid-sized companies.
I have built multi-agent systems in a large enterprise and knowledge assistants for organizations without an IT department. I know what matters where.
- 04
I stay.
After go-live, I take over operation and further development. Everything is documented, so your project never depends on one person.
06 — My stance on AI
Companies that use AI well will get faster. The others will struggle.But using it only because everyone else does burns budget.
What that means in practice
Not every process needs an AI agent. Sometimes a simple automation is the better solution – and then I’ll tell you.
07 — Commitments
What you can rely on.
- 01
You always talk to me.
No sales layer, no handovers. The person who writes the proposal builds the system and looks after it afterwards.
- 02
You know the price before we start.
I deliver clearly scoped projects at a fixed price. If the scope changes, we talk about it first – not on the invoice.
- 03
You see something real early on.
A prototype with your real data instead of a slide deck. So you know whether it works before the big budget is spent.
- 04
Code and data belong to you.
Source code, access and documentation stay with you. You never depend on me, even though I’m happy to stay.
08 — About me
Economist, auditor, CTO. Now an AI Engineer.
At school, I had a failing grade in computer science. I learned to code at 26, at night, alongside my job at KPMG. In 2020 I quit, started as a junior developer and was a CTO a few years later. Today I build AI systems – and I still look at every project through the eyes of the business first.
M.Sc. Economics (LMU) · Ex-KPMG · Ex-CTO · Software since 2020 · Stuttgart, remote
My storyWhat does an AI project cost?
That depends on data sources, integrations and quality requirements. Before you invest more, you get a fixed price after the Process Check – and a calculation of whether it pays off.
How long does it take?
A first pilot with real data is often ready after a few weeks. How long it takes to reach a production system depends on integrations and permissions. We clarify that in the Process Check, before you commit.
Is it worth it for us?
That’s exactly what the Process Check answers: What does the process cost you today, what would the solution cost, and does it pay off? At the end, you get a clear go or no-go. Sometimes the honest answer is that a simple automation is enough.
Can you handle a larger project on your own?
Yes. I built a multi-agent system for an automotive group and delivered knowledge assistants on my own, from requirements to operation. You have one fixed contact person for the whole system, and everything is documented so your project never depends on one person.
How do you handle data protection?
Data protection is part of the architecture, not an afterthought. Where needed, hosting and models run in the EU, the AI only sees what the respective user is allowed to see, and your data is not used to train third-party models.
10 — Start a project
Let’s find out what AI can do in your company.
Tell me in two or three sentences what it’s about. You’ll get an honest assessment within one business day – even if it says you don’t need AI at all.